Scholarlink

<p>Identifying overlapping communities in complex networks is an important aspect of network analysis. This task, however, becomes extremely difficult when working with large-scale directed weighted networks, such as social media networks like Facebook, X, LinkedIn, biological systems, collaboration networks, neural networks, etc., which may consist of millions of entities. There are very few algorithms available for this task, and most are merely extensions of well-known algorithms designed for undirected networks. Besides, these algorithms often produce disjoint and low-quality community structures. On the other hand, evaluating overlapping community structures in directed weighted networks is also a major challenge due to the lack of specific quality measure designed for such networks. To address these challenges, we developed an algorithm called Community Detection in Directed And Weighted Networks (CD-DAWN). It is designed to identify overlapping communities in large, complex networks by selecting and expanding seeds, without needing any prior knowledge about the communities. We also developed a directed weighted overlapping modularity (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(Q_\text {dwo}\)</EquationSource> </InlineEquation>) to evaluate overlapping community structures in directed weighted networks, which is the first such measure to the best of our knowledge. To evaluate the effectiveness of CD-DAWN, we performed extensive experiments on both real-world and artificial networks. The results demonstrate that the proposed approach outperforms the baseline algorithms, accurately detecting high-quality and stable overlapping communities in complex networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Overlapping community detection with a new modularity measure in directed weighted networks

  • 期刊论文

  • 发表日期：

    2025年8月27日

查看全文
  • Abhinav Kumar,
  • Anjali Kumari,
  • Pawan Kumar,
  • Ravins Dohare

摘要

Identifying overlapping communities in complex networks is an important aspect of network analysis. This task, however, becomes extremely difficult when working with large-scale directed weighted networks, such as social media networks like Facebook, X, LinkedIn, biological systems, collaboration networks, neural networks, etc., which may consist of millions of entities. There are very few algorithms available for this task, and most are merely extensions of well-known algorithms designed for undirected networks. Besides, these algorithms often produce disjoint and low-quality community structures. On the other hand, evaluating overlapping community structures in directed weighted networks is also a major challenge due to the lack of specific quality measure designed for such networks. To address these challenges, we developed an algorithm called Community Detection in Directed And Weighted Networks (CD-DAWN). It is designed to identify overlapping communities in large, complex networks by selecting and expanding seeds, without needing any prior knowledge about the communities. We also developed a directed weighted overlapping modularity ( \(Q_\text {dwo}\) ) to evaluate overlapping community structures in directed weighted networks, which is the first such measure to the best of our knowledge. To evaluate the effectiveness of CD-DAWN, we performed extensive experiments on both real-world and artificial networks. The results demonstrate that the proposed approach outperforms the baseline algorithms, accurately detecting high-quality and stable overlapping communities in complex networks.